Replication Server uses a simple caching mechanism to optimize I/O. This mechanism reduces write latency and improves reader speed, since data can usually be read quickly from the cache.
A cache is made up of multiple pages and each page is made up of multiple adjoining blocks. A cache is allocated for each queue at start-up time. Changing the page size changes the size of I/O in the stable queue devices. When a page is full, the entire page is written in one single write operation.
In stable queue caching, the page pointer moves forward and rotates back at the end of the cache. SQM flushes the current page if the writer has filled the message queue and is blocked when waiting for messages. Only blocks with data are written to a disk when flushing a page that is not full.